Developing technical and aesthetic precision that defines serious work
Treat AI as content, not magic
The first time I tried a steady stream, I felt like a magician with a sleight of hand. Every gesture revealed something amazing and utterly wrong — extra fingers, melting eyes, light that refused to obey physics. I laughed, then flared, then stayed until three of me trying to “fix” it.
That’s how I learned the hard way: AI is not magic. It’s material – stubborn, interesting and occasionally uncooperative. Like clay that changes texture every ten minutes.
Artists who are the last in the field stop chasing wonder and start studying friction. Rafik Anadol doesn’t talk about “AI imagination” – he talks about data choreography. Helena Sarin treats her models like pigments, coaxing painterly warmth out of the cold code. Sofia Crespo explores the datasets as naturalist insect studies.
Of course it is mysterious. (Every time I tell myself I understand how diffusion models interpret structure, they prove me wrong.) But mystery does not mean miracle. It means you’re paying attention.
Anyway – if the picture looks too perfect, it probably isn’t.
The importance of repetition and rejection
If AI art has taught me one thing, it’s the art of letting go.
Once, I created about 900 variations of the same theme – a figure dissolves into a geometric blur. By the 700 mark, I stopped feeling smart. By the 800 number, I wanted to throw my computer out the window. Somewhere around 850, the right one appeared. Ironic: It matched the other image of the day.
This is repetition. Not glamorous, but necessary.
Artists like to talk about “quick engineering,” but what we’re really doing is cultivating a relationship with failure. At a workshop last year, one participant admitted that she deletes 95% of her results. The rest she names, edits, archives and sometimes works on months later. “I’m not much,” he said. “I’m stubborn.”
There’s a reason traditional artists practice scales, or practice the same brushstrokes for weeks. Repetition sharpens intuition. It also forces you to admit when something doesn’t work out—which, frankly, happens most of the time.
Isn’t it surprising that the more we automate, the more patience we need?
Quality control and professional standards
The romance of experience only gets you so far. Finally, the art world will ask: Can it hold up under scrutiny?
A digital file, no matter how poetic, should still meet professional standards. Resolution metadata. Color profiles. Consistency These things fade until you lose a print because you forget to track the seed parameter. (Been there. Painful.)
Curators have started raising the bar. At a panel at Serpentine Galleries earlier this year, one noted that AI art “can’t rely on novelty forever — it has to be as technical as any medium.” This means knowing your dataset, understanding model bias, and managing your archive.
It also means taking care of the hidden layers: how the file is named, how reproducible it is, whether it will exist ten years from now. I once spent an afternoon trying to reopen a “ready-made” piece, only to discover that the model version had been deprecated. The image was gone for good. Like losing a painting to mold, except the mold was an update.
Professionalism in AI art is not bureaucratic. It is devotional.
How to build an iterative workflow
When I tell people I keep a spreadsheet of my tips, they laugh. Then they see it and stop laughing.
Each row contains a model version, badge, date, concept note, and one or two adjectives that describe the mode. It is painful. It is also my safety net. Without it, I’d never remember which combination of parameters produced that perfect cyan haze or that amazing, cinematic glow.
An iterative workflow doesn’t eliminate automation—it preserves it. Rafik Anadol calls his practice “data choreography” for a reason: structure gives improvisational meaning.
My own ritual begins with writing. Just a few sentences about what I’m trying to feel, don’t see. . If I’m lucky, the results line up with my mean. Most days, they don’t — but this analogy teaches me something, too.
Some days, the process feels like therapy. Others, such as filing taxes. Both are probably necessary.
Why craftsmanship still matters in digital art
It’s a temptation to think that AI has made craft irrelevant – that art is no longer a matter of creation but of curation. I don’t buy it.
Craftsmanship is the muscle that vaporizes meaning. It’s the difference between a fast-moving visual phenomenon and a lasting artistic language.
Think of the controversy in 2023 when an AI-filled image in Colorado won a regional art prize. The debate wasn’t really about “authenticity.” It was about effort. The winning task felt arbitrary, like a lucky draw. Crafting isn’t about controlling the machine – it’s about owning the choice.
The paradox is that in this age when technology makes everything easy, the effort It has become the new luxury. The labor itself—the visible trace of time, care, and revision—is what gives digital art its weight.
I sometimes wonder if a machine knows when it has worked hard. Probably not. But we do.
A little way
Last week, I was exporting a series of AI filled prints when my power flickered. The screen went black. I sat there, waiting for the GPU to get back to us. In that interval, the room felt almost analog—quiet, smelling faintly of dust, coffee, and ozone.
When the lights came back on, the render was ruined. I breathed, laughed, and started again. The second version was better. Maybe the machine just needs a break. Or maybe I did.
However, perfection has been eliminated.
Craftsmanship won’t save us from automation, but it will keep us honest. The next generation of digital artists will inherit tools that make today’s models seem prehistoric. The challenge will not be learning the technology – it will be learning through it itself.
AI can produce limitless images, but only artists can create limited meaning – select, deliberate, care for.
The craft is a moving target. Some days it feels like a skill. Others, such as a conversation with a machine that refuses to listen.
Maybe that’s right. Maybe it’s work.
Jean-Marie Bonathos (as JM Bonathos Publishing) is the author of two books on digital art:
How to Thrive in the Digital Art Market: An Artist’s Guide to Online Platforms, Collectors, and Trends
The Digital Artist’s Guide to Success
He has also authored seven books on the human side of artificial intelligence and seven other books on non-fiction filmmaking. His writing also explores the intersections of AI, digital art, and nonfiction filmmaking. Check out his latest books: www.jmbonthous.com
He also blogged about the human side of AI, in a separate blog, on Medium: https://medium.com/@jm_26203
Connect with him on Medium to stay in the loop with the latest stories about digital art and @jmbonthous1 to stay in the loop with the latest stories about the human side of AI.